Exploring carbon futures in the EU power sector
Using Exploratory System Dynamics Modelling and Analysis
to explore policy regimes under deep uncertainty
Erwin Loonen, Erik Pruyt, Caner Hamarat
Delft University of Technology
Faculty of Technology Policy and Management
Abstract
The European Emissions Trading Scheme (ETS) in combination with other renewable electricity (RES-E)
support schemes such as (premium) feed-in tariffs or tradable green certificates do not guarantee a carbon
neutral power sector in 2050. This paper shows that many plausible futures of high carbon emissions exist
when no substantial efficiency measures are taken in high growth futures. Using System Dynamics (SD) in
combination with Exploratory Modelling and Analysis (EMA), it seems that the main European energy policies
might result in high levels of carbon abatement but have very limited guarantees whatsoever. There are
potential ‘free lunches’ for policy makers to reduce carbon emissions but these will probably not suffice when
ambition levels remain high. This paper sheds new light on the path to find policy synergies for the European
electricity sector with the aim to rule out lurking catastrophic futures of high carbon emissions combined with
high costs for society.
Keywords: Carbon emissions; Deep uncertainty; Energy transition; EU power sector; ESDMA; Robust policy
design
1. Introduction 1.1 Targets for a sustainable energy sector in the EU
This section introduces the problem related to the design With the studies on the potential environmental and
of policies that will result in a reduction of carbon impacts of gr h gas due to
emissions in the EU power sector. combustion of fossil fuels [1-4], political support was
created to combat climate change in the European Union
(EU). Therefore, reducing greenhouse gas emissions is
one of the 3 main objectives for EU energy policy design.
Abbreviations: ABM, Agent Based Modelling; CCS, Carbon The other two objectives are securing energy supply
Capture and Storage; CP, Copper Plate (model); ETS, Emissions (decreasing the dependency on imported hydrocarbons)
Trading Scheme; ECR, European Climate: Fund; EG European and increasing competitiveness (reducing price rises and
Commission; EU, European Union; EMA, Exploratory Modelling creating growth'and jobs) [5]
and Analysis; ESDMA, Exploratory System Dynamics Modelling
and Analysis; FS, Feature Selection; FIT, Feed-in Tariff;
GW(h), Giga Watt hour; IGCC, Integrated Gasification Combined
Cycle; IPCC, Intergovernmental Panel on Climate Change; IEA,
International Energy Agency; KDE, Kernel Density Estimation;
In an attempt to reduce carbon emissions and increase
the share of renewables in the power mix the European
Commission (EC) launched the European Emissions
LCOE, Levelised costs of electricity; MIC, Marginal Investment Trading Scheme (ETS) and Directive 2009/28/EC on the
costs; MW(h), Mega Watt (hour); MS, Member State; NGCC, promotion of renewable energy sources (RES) [6, 7].
Natural gas combined cycle; PRIM, Patient Rule Induction This directive aims to achieve 20% carbon emissions
Method; RB, Regional Blocks (model); RES-E, Renewable reduction (compared to 1990) and a share of 20% RES.
Energy Sources for Electricity Generation; SD, System The power sector needs to contribute substantially to
Dynamics; TW(h), Terawatt (hour) achieve this target, aiming at 35% renewables in 2020.
Longer term ambitions for carbon emissions reduction in
the power sector are even higher. In 2009, the European
heads of state signed a declaration to reduce carbon
emissions by 80-95% in 2050 [8]. And the power sector
itself has committed to become practically carbon
neutral in 2050 [9].
Next to the just-mentioned directive on renewable
energy supply, the European Commission (EC) launched
the European Emissions Trading Scheme (ETS) in 2005
[7]. The ETS system is a market based instrument to cap-
and-trade carbon emissions allowances amongst large
polluters in Europe. In 2013 phase 3 of the ETS system
commences. From then, power producers have to buy
carbon allowances from the market to compensate for
their emissions. The annual cap of carbon decreases
linearly to trigger an increase in carbon abatement.
When large emitters fail to comply with the ETS system,
they have to pay a penalty which is currently EUR
100/ton CO. This system will be in operation until at
least 2020 but probably longer [7, 10].
Looking at the overall achievements of the EU economy
to reduce carbon emissions, we see that about 15%
emissions reduction is achieved in 2010 compared to
1990. It is interesting to see that most reduction was
achieved during the economic downturn in 2007-2009.
However, the (short) economic recovery in 2010 and a
relatively cold winter in caused a 2.4% increase in
carbon emissions again [11]. On the other hand,
renewable electricity investments are lagging further
behind. Despite increasing efforts, the interim target of
21% electricity from renewable sources (RES-E) in 2010
was missed by 2%. So, a steep increase in EU wide
investments in low carbon technologies would be
needed to achieve these targets but is unlikely to happen
ina short to medium term [12].
1.2 Characteristics of the EU power sector
Most of the electricity in the EU is generated in large
centralized thermal power plants. This makes the power
sector a potential interesting sector for large scale cuts in
carbon emissions. As an interpretation of “Trias
Energetica” [13], carbon emissions reduction in the
power sector can either be achieved by fuel switching,
demand reduction, fossil fuel efficiency gains and/or
carbon storage and sequestration (CCS).
In order to study technological change and the
performance on carbon emissions reduction in the EU
power sector, a micro-economic perspective is needed
[14]. Private investments mainly shape the power mix of
different generation technologies in the current
liberalized power market.
Dynamics in the power sector, such as the just
mentioned investment dynamics, are characterized by
long construction lead times, permit lead times and
lifetimes of technol Furthermore, there
are many interactions, delays and feedbacks in the
system that can cause cyclic behaviour, lock-in effects
and path dependency [15]. This implies that a long term
modelling perspective is needed to study the effects of
these dynamics.
When studying the EU power sector and its long term
dynamics, many uncertainties can be identified that that
influence the system and shape its future. Examples of
uncertainties can be ‘hard’ parametric values like
installed capacities and economic lifetimes but also ‘soft’
values, such as weight factors on strategic motives of
private investors. Besides these parametric
uncertainties, also structural uncertainties exist, such as
the effect of low capacity reserve margins on strategic
market bidding and the effect of demand growth on
investment forecasting [16]. The level of uncertainty
dealt with in this study is called “deep uncertainty”.
According to Lempert et al. [17]: “Deep uncertainty
exists when analysts do not know, or the parties to a
decision cannot agree on, (1) the appropriate models to
describe the interactions, among a system's variables,
(2) the probability distributions to represent uncertainty
about key variables and parameters in the models,
and/or (3) how to value the desirability of alternative
outcomes”.
1.3 Implications for policy analysis and design
In order to effectively design policies that contribute to
the objectives of a safe, secure and sustainable electricity
supply in the EU, the earlier mentioned characteristics of
the power sector imply that policy makers should focus
on long term dynamics and influencing private
investments. Therefore, this study uses a timeframe of
the next 40 years and a uses micro-economic approach
for modelling investment dynamics.
Given the many uncertainties that will shape the future
EU power sector, policies are needed that will suffice in
all ‘plausible futures’. This is called robust decision
making [17] and asks for a research method that is
capable to incorporate both the complex dynamics as
well as the uncertainties, as mentioned in 1.2.
Exploratory System Dynamics and Analysis (ESDMA) is
such a method. Section 2 provides a short explanation of
ESDMA.
1.4 Previous research on policy design to decarbonize the
power sector
A literature study was performed on carbon emissions
reduction and policy design studies in the power sector.
This included system dynamics modelling studies [15,
18-24], agent based modelling studies [25], cost-
efficiency optimization studies [9, 26, 27], and ex-post
policy assessment studies based on empirical and
literature research [28-30].
What all these studies have in common (although some
more than others) is that they fail to incorporate the
effects of deep uncertainty in their analyses. Therefore
the aim of this study is to incorporate deep uncertainty
in the analysis of a transition towards a low carbon EU
power sector, and suggest directions for policy design
that provide satisfying outcomes in all plausible futures,
as suggested by [17].
1.5 Research question
The just mentioned aim of this study leads to a desire to
explore the impact of uncertainties on plausible carbon
emissions futures in the power sector. The following
research question fits this aim:
What are the effects of parametric and structural
uncertainties on plausible carbon emissions futures
in the EU power sector under different policy
regimes to provide directions for robust policy
design?
The study on the EU power sector presented here aims
to (1) explore and not predict a wide array of plausible
futures for carbon emissions reduction, (2) identify most
risky and promising futures for decarbonisation, (3)
assess the robustness of different policy regimes under
different uncertainties and their plausible ranges, and
(4) provide directions for policy making to enhance the
robustness of policy regimes under review.
The next section indicates the policy regimes under
review.
1. 6 Policy regimes under review
Given the dynamic and evolving character of the EU
power sector's policy landscape, different policy regimes
will be explored. A selection is made of 4 different
regimes. Besides the current ETS system, one reference
regime without the ETS system is also explored (i.e. No
ETS). Furthermore, to two regimes are explored where
the ETS system co-exists with a renewable electricity
support scheme, these are a premium Feed-In Tariff
(FIT) system and a Tradable Green Certificates (TGC)
system.
e No ETS - A free and competitive market
without the ETS system or support policies.
e ETS Only - The ETS system is introduced with
different potential carbon cap pathways from
2020.
e FIT - On top of the ETS system, a premium
feed-in tariff (FIT) is introduced that covers the
gap between the electricity market price and
the lowest levelised costs of electricity (LCOE)
of some RES-E technologies'. An additional
profit mark-up of 100% of is added and the
tariff is limited to a maximum of EUR
150/MWh.
« TGC - On top of the ETS system, an EU wide
suppliers’ obligation to buy Tradable Green
Certificates (TGC) is introduced. The obligation
increases towards 80% fraction RES-E in 2050
and non-discriminatory for all RES-E
technologies.
All policy regimes act in an energy-only market. Energy-
only markets rely on the electricity market to provide
investment incentives for power generation capacity
[32]. This means that no fees are paid for installed
capacity or other types of capacity mechanisms are in
place. Loonen [16] provides a more detailed elaboration
of the policy regimes explored in this study.
It has to be noted that this study does not aim to assess
the effectiveness of the policy regimes for comparison.
Tools and probabilities are lacking in this study to make
such a comparison useful. The mere aim is exploration
only.
1 RES-E technologies that fall under the premium
FIT policy regime are wind power, biomass and PV
solar power. Large scale hydro power is not
included for two reasons: (1) Most FIT systems
currently applied in the EU Member States mostly
only account for wind, biomass and/or PV solar
power [12], and (2) generating costs of large scale
hydro power is generally substantially lower than
the other RES-E technologies [31] so including large
scale hydro power undermines the effectiveness of
this policy support regime. The feed-in tariffs are
regional specific in the Regional Blocks model.
1.7 Organization
The structure of this paper is as follows: Section 2
discusses Exploratory System Dynamics Modelling and
Analysis as the research method for this study; in section
3, the results from the analysis on the base policy
regimes are presented and interpreted; section 4 shows
the results of an assessment of 4 policy directions on
their robustness; in section 5, conclusions of this study
are stated; and in section 6, a discussion on the validity
of the outcomes and ex-post criticism is presented.
2. Modelling policy issues for exploratory purposes
After the introduction of the problem and aim of this
study, the steps taken in modelling for exploratory
purposes are listed. This section elaborates on
Exploratory Modelling and Analysis (EMA) in
combination with System Dynamics (SD) models as main
research methodology.
2.1 Methodology
In order to explore the behaviour of the EU power sector
in a wide array of plausible futures, Exploratory
Modelling and Analysis (EMA) can be used [17, 33, 34].
EMA involves the design of plausible models and
identification of most important uncertainties and their
plausible ranges in order to generate plausible futures.
These uncertainties and ranges are used as input for the
simulation models.
An approach that builds forward on EMA, focussing on
policy design under deep uncertainty is the Adaptive
Robust Design (ARD) approach [35]. The steps of the
ARD approach are used in this study.
Steps in ARD are:
1. Conceptualize the policy problem
2. Specify the uncertainties relevant for policy
analysis
3. Develop an ensemble of models for exploring
uncertainties
4. Run the computer models without any policies
in order to generate the ensemble of futures
5. Explore and analyse the results from step 4 in
order to identify the troublesome and promising
regions across the outcomes of interest, as well
as the main causes underlying these regions
6. Design candidate policies for addressing
vulnerabilities and seizing opportunities
7. Implement and test the candidate policies across
the ensemble of futures
8. Iterate through steps 5-7 until satisfying policies
emerges
2.2 ESDMA and incorporating uncertainty
EMA asks for simulation models as the experimental
setup to generate and explore plausible futures of a
system under research. Although many different
simulation models could be used, in this study EMA is
combined with System Dynamics (SD) models. System
dynamics models are particularly able to incorporate the
characteristics of the power sector as mentioned in
section 1.2. Combining EMA with SD modelling is called
ESDMA [36, 37].
The second step in ARD is to identify uncertainties that
are relevant for the problem under research. In this
study, all uncertainties are divided in parametric and
structural uncertainties.
Parametric uncertainties are values of parameters
concerning relationships in the system [33] and are
defined by a single value. These uncertain values are
constant in during a whole simulation run. Examples of
parametric uncertainties are initial values (eg.
generation capacities), constants (e.g. economic lifetimes
of technologies), delays (e.g. information delays to
forecast expected future electricity demand), and
switches (e.g. to turn on/off a part of the model
structure).
Structural uncertainties indicate specific structures of
the model that can be turned on/off. By incorporating
structural uncertainties in the model, the ability exist to
use different assumptions about factors and
relationships in the model [33]. Examples of structural
uncertainties are: The effect of economic growth and
electrification on electricity demand, the availability of
battery storage to cope with intermittent electricity
supply and the effect of different investor's perspectives
on capacity investments for future electricity supply.
A wide range of literature sources is used for identifying
most important uncertainties and their plausible ranges
[31, 38-53]. Furthermore, a workshop and interviews
were held with scholars and experts from the power
sector to draw potential evolutions of some of the main
structural uncertainties like economic growth,
electrification rates and battery storage in electric
vehicles. Moreover, the most important strategic motives
that influence investment decisions were also verified
during this workshop. However, when no real world data
could be obtained, guestimates were made. All
uncertainties and the way they affect the simulation
models in the study can be found in [16]. The
uncertainties and their ranges are listed in a Python [54]
shell that is connected to the simulation models in order
to provide the input values for the models.
2.3 Tools for exploration
Thousands of scenarios can be generated by connecting
the SD models to a Python shell that ranges the input
values for each run. When a dataset is generated of all
runs, each ensemble of runs needs to be explored.
Different tools exist to explore the large amount of
scenarios. Tool used here are:
e Explorative visualization by means of envelopes
that show the upper and lower limits of the
scenario ensemble over time. These envelops are
complemented with a kernel density estimation
(KDE) on the end-state of each run [55, 56].
e The Feature Selection (FS) algorithm to identify
uncertainties that have the largest influence on a
specific performance indicator (PI) [57].
Patient Rule Induction Method (PRIM) to identify
combinations of uncertainties that are highly
predictive for a specific model output [58].
2.4 Simulation models
As mentioned before, ESDMA needs plausible simulation
models for exploration. In this study, 2 different
simulation models of the EU power sector are designed.
Most important distinction between both models is the
modelling aggregation level. These 2 models are called
the Copper Plate model (highly aggregated) and the
Regional Blocks model (more detailed model, see figure
2). However, the basic structure of both models is largely
the same. In total 9 different technologies are included
that compete for investments (i.e. coal, gas, biomass, PV
solar, hydro, nuclear, wind, gas with CCS and coal with
CCS). The specific technological characteristics and the
total uncertainty space is found in [16].
Figure 1: Schematic representation of the EU Regional
Blocks model. This model includes regional characteristics
for renewable electricity supply and interconnection
capacity limitations for power trade and supply.
In a liberalized power market commercial parties
determine the type and capacity amount of generation
technologies. A micro-economic perspective is needed in
order to research investor’s behaviour in a liberalized
power sector, such as the EU [14].
Figure 1 shows the main dynamics that underlie the
simulation models used in this study. At a high
aggregation level, basically 2 important factors drive
new capacity investments. These are profitability and
electricity demand. Investors assess the profitability of
each technology, in order to determine the amount of
new capacity investments. These dynamics and model
structures are further elaborated in [16]. The levelised
costs of electricity (LCOE) indicate the total costs during
the whole ic lifetime of a technology, divided by
the total (expected) power generation during its lifetime.
The definition and formula of LCOE used in this thesis
study is stated in [16].
transmission and *
distribution fosses
Electricity demand
‘eketifcation rate
‘economic growth
proportionality constant
industrial carbon
emissions —~
_~_-
ETS emisions cap avaibbity factor
x.
we
\
demand price
chasticity fictor
fel efficiency price cap
Installed capacity
carbon emissions
penal
rest of world
resource demand
>| Levelised Costs of
Elect
reduction rate
<——~ ina! marginal
ee investment costs Electricity price
ns
construction kad
resource price
os! elasticity factor
n of fire
rest of worl! fie]
demand
fuel price elasticity
factor technology
progress ratio
EV battery storage
SO, econcimic tine
\x
fixed O&M costs
Profitability
Figure 2: Highly aggregated causal loop diagram of investment dynamics in power generation capacity with main uncertainties,
from a micro-economic perspective [14, 48].
3. Exploring base policy regimes
3.1 Visual inspection of envelopes with kernel density
estimations (KDEs)
Although the primary focus of this study is on carbon
emissions, other performance indicators are used to put
the performance into context. These performance
indicators are related to production, policy and
investment costs, as well as renewable electricity supply.
Outcomes on all these performance indicators are
presented in [16] but only most relevant outcomes are
presented in this paper.
3.2 Carbon emissions in base ensembles - Risks and
opportunities for a low carbon power sector
Looking at the envelopes on carbon emissions (figures 3-
5), a wide distribution of plausible end-states in 2050 is
seen. It is not surprising to see that most risk prone high
carbon futures are seen in the No ETS policy regime. In
the Looking at the other policy regimes, worst high
carbon futures are less severe but still rather bad.
From these envelopes, it seems as if introducing the ETS
system, with or without a RES-E support scheme, is a
push in the right direction to reduce carbon emissions
but is not enough to effectively rule out all worst case
futures.
Besides these catastrophic futures, also very low carbon
promising futures are observed in all base policy
regimes. However, most promising futures that lead to
an almost carbon neutral power sector in 2050 are
observed in the policy regimes where the ETS system co-
exists with a renewable electricity support scheme (i.e.
the FIT or TGC regime). From these observations, it
seems possible to decarbonize the EU power sector to a
large extent when opportunities are seized but there is
no base policy regime that guarantees large emissions
reductions.
blue =No ETS
Green = ETS only Purple = 16C
Regional Blocks
Copper Piate
25ea0
25EH0
Figure 3: Annual carbon emissions envelopes base policy regimes [ton/year], 10.000 runs. High carbon futures are plausible in
all policy regimes.
reen = ETS only Purple = TSC
Regional Blocks [EEE HE] Sues NOeTS Red = FIT
Gr
25611
Figure 4: Ci carbon
futures are observed in all policy regimes
Green = ETS only Purple = TGC
Regional Blocks yt] BMS AGES
2010
25EH2
pes base policy regimes [ton], 10.000 runs. A wide range of plausible carbon
Figure 5: Average carbon emissions envelopes base policy regimes [ton/MWh], 10.000 runs. A trend towards lower average
carbon emissions is observed in most policy regimes.
When looking at the financial implications of each policy
regime (figures 6-8), it seems that for most of the
promising low carbon futures the implications can be
significant for society (in case of FIT), end-consumer (in
case of TGC) and producers (in case of FIT and TGC).
Although we do not exactly know the correlation
between the high costs futures and the high carbon
reduction futures in the TGC and FIT policy regimes, the
envelopes of the No ETS and ETS policy regimes show
that there may be some opportunities for ‘free lunch
policies’ available. ‘Free lunch’ policies are policies “that
improve some or most measures of performance without
degrading others” [59]. In this study it means that
carbon emission can be reduced without increasing costs
for society or producers. Reason is that in the No ETS
and ETS only policy regimes some futures are observed
where carbon emissions are reduced significantly, while
the total costs do not necessarily increase significant.
Copper Plate Blue=ETS only Red = TGC
Green = FIT
2o10 “ 2050
Figure 6: Average costs? of policies borne by society
and/or end-consumer due to the ETS system and RES-E
support schemes, Copper Plate model 10.000 runs.
Copper Plate Tam SE] Blue=NOETS Red= FIT
Green=ETS only Purple = TGC
C)
do10 * ° . * 2050"
Figure 7: Envelopes of average costs for producers base
policy regimes [EUR/MWh], Copper Plate model 10.000
runs.
Copper Plate Blue=NoETS Red = FIT
ssa Green =ETS only Purple =T6C
300
——— —= \
é = =
2010 “ - 2050
Figure 8: Average total costs envelopes base policy
regimes [EUR/MWh], Copper Plate model 10.000 runs.
Total costs consists of policy costs added with costs for
producers
2 These costs are averaged out over the total run-time.
Negative costs indicate a revenue stream for policy
makers/society, due to ETS allowances auctioning.
3.2 Identifying influential uncertainties on carbon
emissions and costs
To identify the main causes and uncertainties underlying
high and low carbon futures, two methods are used as
introduced in section 2.3. These are a Feature Selection
(FS) algorithm and the Patient Rule Induction Method
(PRIM).
Tables 1 and 2 show the relative scores of individual
uncertainties by means of the FS algorithm on annual
carbon emissions. This is based on an underlying
classification scheme; higher the annual carbon
emissions reduction yields higher scores. These scores
should merely be interpreted in a relative way and not in
an absolute way.
It is not very surprising that economic growth and the
rate of electrification of the economy turn out to have a
high influence on annual carbon emissions in all policy
regimes. The reason is that these uncertainties drive
electricity demand and subsequently supply. In a power
sector that largely consists of conventional high carbon
emitting power plants, it seems obvious that these
uncertainties are most influential. On the other hand, the
extent of the influence of these uncertainties may be
surprising, compared to the other uncertainties.
Furthermore, it seems that the relative influence of
economic growth and electrification decrease with
introducing the ETS system and a RES-E support scheme
(FIT or TGC). This also makes sense because these policy
schemes intent to make a transition in the power mix
from conventional high carbon emitting technologies to
low carbon and renewable technologies. This is caused
by the underlying mechanism of profitability that drives
new investments. The ETS and RES-E support schemes
intent to reduce the profitability of high carbon
technologies while increasing the profitability of
renewable and low carbon technologies. This increases
the relative attractiveness of low carbon renewable
technologies for investors.
Moreover, facilitating higher penetration levels of wind
and PV solar power (intermittent RES-E technologies)
seems to be a condition to further reduce carbon
emissions. This could be done by using batteries in
electric vehicles (or other storage possibilities) to
temporarily store electricity in order to deal with the
variability of supply.
Table 1: Relative feature selection scores on carbon reduction Regional Blocks model base policy regimes. Economic growth and
electrification of the economy yield highest scores.
Feature No ETS ETS Only FIT TGC
Economic growth 0.207 0.158 0.142 0.157
Electrification rate 0.254 0.178 0.162 0.111
ETS price determination 0.015 0.005 0.031 0.058
Interconnection capacity expansion 0.030
Storage for intermittent supply 0.002 0.028 0.030 0.035
Investors’ overinvestment factor 0.001 0.042
Table 2: Relative feature selection scores on carbon reduction Copper Plate model base policy regimes. Economic growth and
electrification of the economy yield highest scores.
Feature
Economic growth
Electrification rate
Availability factor
Physical limitations large scale hydro
Storage for intermittent supply
The FS algorithm is also performed on the average total
costs for electricity (this is the total costs of policies added
to the total costs for producers and averaged out over the
electricity generated during the period 2010-2050). The
FS algorithm returned again economic growth and
electrification rate for all policy regimes as important
drivers for high costs, but their relative influence is less
compared to the FS analysis on carbon emissions [16]..
However, it seems that both from a carbon emissions
perspective and costs perspective, the main drivers of
increasing electricity demand should be targeted with
policy measures.
Some uncertainties are only highly influential in
combination with others. Together, these can create risks
or opportunities. PRIM is a method to identify these
combinations. This method is performed on all
performance indicators, shown in [16], but only a brief
summary of the most interesting outcomes is presented
here.
Comparable to the outcomes of the FS analysis, the PRIM
analysis on annual carbon emission returned economic
No ETS
ETS Only FIT TGC
0.315 0.249 0.236 0.147
0.248 0.192 0.183 0.186
0.043 0.010 0.001
0.042 0.021 0.042 0.035
0.041
growth and electrification rate as most influential.
However, the analysis on average carbon emissions per
MWh showed some other interesting outcomes. When
choosing a No ETS policy regime, worst cases can be ruled
out by reducing the carbon intensity rate of fossil fuel
generation technologies, and by exploiting the full growth
potential of large scale hydro power?.
When a TGC system is chosen, some additional interesting
results are seen in the PRIM analysis. At first, fuel price
elasticity factors counteract a transition towards a
renewable electricity supply. This means that when
demand for fossil fuels decrease, prices decrease
proportionally which make fossil fuels more attractive
again. So in case of a large transition towards renewables,
3 Different estimates exist on the potential of large scale
hydro power in Europe [60, 61]. This is amongst others
dependent on what is perceived as acceptable potential for
society and environment. This study does not elaborate on
potential consequences for society and environment to
increase the potential for large scale hydro power.
10
policy makers could counteract this effect by assuring
structurally high fossil fuel prices (e.g. by increasing taxes
or levies).
Furthermore, too optimistic estimations of future demand
growth by investors also decrease average carbon
emissions. Reason is that most renewables (except
biomass) benefit from their low variable costs which give
them priority position in the merit order for supplying
electricity. So, even when investments are uniformly
distributed over all technologies, excess investments could
lead to lower carbon emissions. However, an
overestimation of future demand growth amongst
investors will also lead to higher average costs for
producers. There are two reasons for this effect. The first
reason is that overinvestments lead to lower average
capacity factors of existing generation capacities, making
their use less efficient. And secondly, overinvestments
could lead to increased resource scarcity which drives
marginal investment costs of new generation capacity.
Last note worth mentioning from the PRIM analysis is that
ambitious short term targets for the ETS and TGC regimes
(i.e. logistic growth or decrease) could have a beneficial
effect on decreasing carbon emissions in the long run.
However, these ambitious targets most likely also increase
the total costs of electricity that needs to be paid by the
end consumer and/or society. So here it seems that a
trade-off needs to be made between reducing carbon
emissions and increasing costs of electricity.
3.3 Conclusions on base policy regimes
The base policy regimes all show a wide distribution in
plausible carbon futures for the EU power sector.
Otherwise than might be expected, the ETS and RES-E
policy schemes assessed here do not guarantee substantial
carbon emissions reductions without additional policy
measures.
There seem to be ‘free lunch’ policies available in reducing
carbon emissions without increasing the total costs of
electricity. These free lunch policies should be searched in
limiting electricity demand growth.
Besides the potential ‘free lunch’ policies, sometimes
trade-offs between costs and carbon emissions reductions
need to be made. These trade-offs could be related to
allowing or stimulating (over)investments in low carbon
technologies. Increasing (over)investments in renewables
could reduce carbon emissions due to the priority position
of most renewables in the merit order. RES-E subsidy
schemes could amplify this effect further. The same
accounts for ambitious short term targets for the ETS and
TGC schemes. When targets are set ambitious in early
phases, higher carbon emissions reduction are seen that
last for the longer term, but this would most probably also
involve higher costs that need to be borne by consumers
and/or society.
Last two observations made for further carbon emissions
reductions, are to allow for higher penetration levels of
hydro power by exploiting its full growth potential. And
secondly, to facilitate (battery) storage of intermittent
electricity supply.
4. Robust policy testing
From the analysis on the base policy regimes and the
analysis on high versus low growth scenarios (see [16]) a
set of new policy measures is suggested and tested here.
These are really suggestions for policy directions and
should not be interpreted as real world policy measures.
4.1 Directions for new policy design
From the previous analyses, we saw that limiting the main
effects on demand increase (economic growth and
electrification rate) seems to be beneficial to reduce
absolute carbon emissions (direction 1). Furthermore, a
condition to allow for large scale penetration of wind and
PV solar power is to make (battery) storage available
(direction 2). Moreover, carbon emissions are most likely
to decrease further with exploiting the full potential of
large scale hydro power (direction 3). And last, in order to
prevent large costs increases for producers (i.e. finally
transferred to the end-consumer), the maximum amount
of overinvestments is limited (direction 4)..
The four directions for new policy design suggested and
tested in this study are:
1. Limit demand growth to max. 1.5% per year.
2. Increase available (battery) storage for
intermittent power supply towards 1.3TW in
2050.
3. Allow for higher penetration levels of large scale
hydro power. See [16] for more details.
4, Limit overinvestments to a maximum of 50% of
the expected future demand gap.
4.2 Exploring new policy ensembles
The directions for policy design in 4.1 effectively rule out
most catastrophic high carbon and high cost futures for all
policy regimes (see figures 9 and 10). Interestingly the FIT
and TGC policy regimes show their best low carbon future
in the original base policy ensembles. This is caused by
policy direction 4 that limits overinvestments. On the
other hand, the sacrifice that is made is relatively
insignificant compared to the progress that is made to rule
out worst high costs futures. Although the total costs
(costs for producers and society summed) are reduced in
most futures, so are the revenues from the ETS system.
Some policy makers could find this undesirable because
this system can be a real cash cow as seen in the base
policy ensembles.
In order to further improve the performance on carbon
emissions, a PRIM analysis was performed again to see
what the most influential uncertainties on carbon
emissions are for the new policy regimes (see [16]). This
analysis indicated that still economic growth and the
electrification rate are most influential in all policy
regimes. This outcome is quite remarkable, given the
policy measure that limits these effects on demand growth
to a maximum of 1.5% growth per year. Apparently,
demand growth remains the most important bottleneck
—="RBno ETS! — RBETSonly! — RBFITI — RBTGCI
— RBnoETS2 — RBETSonly2 RBFIT2 RBTGC2
elt ‘Cumulative carbon emissions
02013 30202023 2030 2088 2040 20830
Time (years) ten
11
for further reducing carbon emissions in the power sector.
However, when the EU has the ambition to become
(almost) carbon neutral by 2050, residual emissions need
to be cut as well by fuel switching or carbon capture and
sequestration. To push such a technological transition, the
results from the PRIM analysis suggest that further
increases in demand reduction should be combined with a
steeply decreasing carbon cap early in phase 4 of the ETS
system (logistic decline) and steeply increasing the
obligation for green certificates short after introduction
(logistic growth). However, following from the results of
the base ensembles, this means that this would
presumably lead to higher average total costs for
electricity.
= thins) No ETS ae
2.5E+11 1.4E+11
Figure 9: Cumulative carbon emissions, base (1 in blue) and new (2 in green) policy ensembles [ton], 6000 runs. Worst case
futures are effectively ruled out with a small sacrifice for best case futures in FIT and TGC.
12
— RBno ETS 1
— RBno ETS2
—" RBETS only 1
—_RBETS only 2
average total costs
RBFIT! = — RBTGC1
—— RBFIT2 = —~ RBTGC2
2010 201520202025 2030.-—«203S_—«2040-—«204S—OAWMDDODD —_0.04528592
Time (years
2010 eo 2050
Figure 10: Average total costs of electricity, base (1 in blue) and new (2 in green) policy ensembles [EUR/MWh], 6000 runs. Total
costs for society and producers largely decreased in worst case futures but so are revenues from the ETS system.
5. Final conclusions
5.1 General conclusions
The European Emissions Trading Scheme (ETS) in
combination with other renewable electricity (RES-E)
support schemes such as a premium feed-in tariff (FIT) or
tradable green certificates system (TGC) certainly do not
guarantee a carbon neutral power sector in 2050 in all
futures. When these systems are implemented in isolation,
there is a great risk of ending up in a high carbon future.
However, the policy directions assessed here seem to be
robust in ruling out most catastrophic future for all policy
regimes.
There are potential ‘free lunches’ policies available to
reduce carbon emissions while not hampering the
performance on other crucial performance indicators. The
‘free lunch’ policies should be directed towards demand
reduction. However, these ‘free lunch’ policies will
probably not suffice when long term ambition levels for
carbon reduction very high as stated by the European
Commission and the power sector itself [8, 9].
Most promising futures are seen when synergies between
the ETS system and RES-E support schemes are yielded.
However, one of the critical success factors is to allow for
high penetration levels of renewables like wind, PV and
hydro power. Wind and PV solar penetration rates could
increase by allowing (temporary) storage of electricity
(e.g. battery storage in electric vehicles).
Furthermore, the balancing effect of lower fossil fuel
prices in a transition towards non-fossil fuel generation
technologies should be counterattacked. One of the
options to might be to introduce taxes (or other levies)
that increase proportional to the decreases in fossil fuel
prices assure structurally high fossil fuel prices. However,
the drawback is that this may interfere with the EC’s
objectives on competitiveness of the EU economy [5] . On
this short term this will cause higher prices for producers
that are transferred to consumers.
The other side of the coin is that there is a risk for high
societal costs when the ETS system coexists with RES-E
support schemes. Limiting risks of substantial demand
growth is again one of the most important measures that
should be taken to limit substantial societal costs, next to
limiting overinvestments.
5.2 Conclusions on policy regimes
No ETS - Energy only market without policies
The No ETS policy regime showed the widest range in
plausible carbon emissions futures. Although some
promising futures are identified, a significant risk to end
up in a high carbon future is taken when choosing this
regime without additional policy measures. The additional
directions for policy design improved the performance
significant, but an increase (even a doubling) in carbon
emissions in this regime remains plausible.
ETS only - Implementing the Emissions Trading
Scheme
With the introduction of the ETS system, the worst case
scenario is far less severe than without the ETS system. On
the other hand, substantial emissions increases are still
very plausible (up to 5 times current emissions) without
additional policy measures. With the given directions for
policy measures tested in this study, risks of high carbon
futures are significantly decreased. However, in worst
cases carbon emissions might still increase up to 50%.
These measures also reduce revenues from carbon
allowance auctioning from a policy maker perspective.
However, electricity consumers will generally benefit if
these lower costs for producers are carried forward in the
electricity price.
(Dynamic) premium feed-in with the ETS system
Like the other policy regimes, the premium feed-in tariff
on top of the ETS system without additional policy
measures does not guarantee high carbon emissions
reduction. However, when the additional measures as
suggested in this study are implemented, the risk of
carbon emissions increase in 2050 is (almost) completely
ruled out. Furthermore, from a policy costs perspective
these additional policy measures are also effective to
prevent the risk of enormous costs for society to finance
this system. The premium feed-in system tested here is
dynamic and not fixed. This can cause the gap between the
electricity market price and LCOE of renewables to
increase, which drives the reinforcing loop that lead to
higher policy costs in order to bridge the gap (until the
maximum feed-in tariff). So, especially demand reduction
and preventing overinvestments are important conditions
to keep costs in hand for society.
Tradable Green Certificates with the ETS system
Comparable to the other policy regimes, a co-introduction
of a TGC system and ETS system as tested here do not
assure carbon emissions reductions towards 2050 without
additional policy measures. However, when the additional
policy measures are implemented, all plausible futures
generated in this study show a guaranteed carbon
emissions reduction towards 2050. And besides that, also
13
an (almost) carbon neutral power sector seems plausible.
Next to carbon emissions reduction the additional
measures tested in this study are also beneficial from a
societal and consumer's costs perspective. Without these
measures, high prices for certificates in combination with
a substantial increasing demand could result in
unacceptable high costs.
The analysis on the TGC policy regime with new policy
measures indicated that in order to further reduce carbon
emissions, limiting demand growth (economic growth and
electrification rate) remains the most important
bottlenecks. And last, ambitious short term targets for
reducing the carbon cap in phase 4 (after 2020) and
increasing green certificates obligations could further
drive carbon emissions reduction. The drawback is
however, that most probably we need to accept higher
costs for our electricity supply in that case.
6. Discussion, criticism and further research
6.1 Plausibility of futures
Plausibility of futures is one of the important conditions
for exploring futures in an EMA study [17]. It is
questionable if whether the demand scenarios sketched in
this thesis study are plausible. A maximum threshold used
in the thesis study of Loonen [16] for plausible power
demand futures is 3TW full load capacity in 2050. It
turned out that a fraction of 7.6-10.3% of all scenarios in
the base ensembles exceeded this threshold. The main risk
of including implausible scenarios in the dataset is that
specific policies are designed to deal with these scenarios.
This risk can be offset by designing adaptive policy
measures rather than static policy measures. Adaptive
policy measures are only activated when a certain
situation occurs that indicates a risk or opportunity to
predefined policy targets. According to Bankes et al. [62],
designing adaptive policies is a very important type of
solution complexity to achieve robustness in an uncertain
future. So, next steps in this research would be related
realistic and adaptive policy design to decarbonize the
power sector, while keeping costs acceptable.
6.2 Ex-post criticism
Sterman [63] stated that “... systems thinking requires
understanding that all models are wrong and humility
about the limitations of our knowledge”. The reality can
never be fully incorporated in simulation models, which is
also true for the study presented here. Some aspects are
intentionally or unintentionally left outside the scope of
research. The aim of this research was to explore the
plausible uncertainty space of future carbon emissions in
14
the EU power sector. However, some humility is required
here.
Possible improvements on modelling and policy design to
decarbonize the EU power sector are:
Extension of the amount of (promising)
generation technologies (eg. geothermal,
concentrated solar power, small scale hydro,
(decentralized) biogas, combined heat and
power, etc.)
More and particularly real world (adaptive)
policy testing. This study included only 1 specific
carbon policy (ETS system) and 2 RES-E support
schemes (FIT and TGC), but there are many more
potential effective policies. Next, only some
directions for policy design are suggested
(demand growth limitation, limiting
overinvestments, etc.) but these lack a sense of
real world policy implementation.
Towards a holistic (carbon emissions in the total
economy) and hybrid (agent based and system
dynamics) modelling approach. Using a holistic
modelling approach allows to explore the
potential effects of shifting carbon emissions
from one sector to another, for example by
electrification. And a hybrid modelling approach
would allow for including the aspects of bounded
rational and discrete event decision making of
agents with the continuous information flows
that influence these decisions [64].
References
1
o
IPCC, Climate Change 2001, Synthesis
Report, Summary for Policy Makers, 2001. p.
4.
IPCC, Climate Change 2007: Synthesis
Report, 2007, Intergovernmental Panel on
Climate Change: Valencia, Spain.
McCarthy, M., Lord Stern on global warming:
It's even worse than I thought, 2009.
Stern, N., Stern Review Report on the
Economics of Climate Change, Executive
Summary, 2006.
European Commission An energy policy for
Europe. 2007.
European Commission, DIRECTIVE
2009/28/EC OF THE EUROPEAN
PARLIAMENT AND OF THE COUNCIL of
23 April 2009, On the promotion of the use of
energy from renewable sources and amending
and subsequently repealing Directives
2001/77/EC and 2003/30/EC. Official Journal
of the European Union, 2009. 140: p. 16-61.
European Commission. Zmissions Trading
System 2012 [cited 2012 31 May]; Available
from:
http://ec.europa.eu/clima/policies/ets/index_en.
htm.
EUR-Lex. International climate policy post-
Copenhagen: Acting now to reinvigorate
global action on climate change. 2010 [cited
2012 1 July]: Available from: http://eur-
lex.europa.eu/Notice.do?mode=dbl&lang=en&i
hmlang=en&lng1=en,nl&lng2=bg,cs,da,de,el,e
n,es,et,fi,fr,hu,it,lt,lv,mt,nlpl,pt,ro,sk,sl,sv,&v
al=509334:cs&page=.
Eurelectric, Power Choices, Pathways to
Carbon-Neutral Electricity in Europe by 2050,
2009: Brussels.
European Commission. Emissions Trading
Scheme, Cap. 2012 [cited 2012 31 May];
Available from:
http://ec.europa.eu/clima/policies/ets/cap/index
_en.htm.
EEA, Annual European Union greenhouse gas
inventory 1990-2010 and inventory report
2012, 2012, EEA: Luxembourg.
Klessmann, C., et al., Status and perspectives
of renewable energy policy and deployment in
the European Union - What is needed to reach
the 2020 targets? Energy Policy, 2011. 39: p.
7637-7657.
Lysen, E.H., The Trias Energetica: Solar
Energy Strategies for Developing Countries,
in urosun Conference1996: ‘Freiburg,
Germany.
Blyth, W., The economics of transition in the
power sector, I.E. Agency, Editor 2010, Oxford
Energy Associates: Paris, France.
Pruyt, E. The EU-25 Power Sector: a System
Dynamics Model of Competing Electricity
Generation Technologies. in 25th
International Conference of the System
Dynamics Society. 2007. Boston, MA.: System
Dynamics Society.
Loonen, E., Exploring Carbon Pathways in
the EU Power Sector, Using Exploratory
System Dynamics Modelling and Analysis to
assess energy policy regimes under deep
uncertainty, in Faculty of Technology, Policy
and Management2012, Delft University of
Technology: Delft.
Lempert, R.J., S.W. Popper, and 8.C. Bankes,
Shaping the Next One Hundred Years, New
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
Methods for Quantitative Long-Term Policy
Analysis2003, Santa Monica: RAND.
Ford, A., System Dynamics and the Electric
Power Industry. System Dynamics Review,
1997. 13: p. 57-85.
Ford, A., Cycles in competitive electricity
markets: a simulation study of the western
United States. Energy Policy, 1999. 27: p. 637-
658.
Ford, A., Waiting for the boom: a simulation
study of power plant construction in
California. Energy Policy, 2001. 29: p. 847-
Ford, A., Simulating the impact of a carbon
market on the electricity system in the
Western USA, in 24th International
Conference of the System Dynamics
Society2006: Nijmegen, The Netherlands. p. 1-
54.
Ford, A., Simulation scenarios for rapid
reduction in carbon dioxide emissions in the
western electricity system. Energy Policy
2008. 36: p. 443-455.
Ford, A., A. Dimitrovski, and K. Tomsovic, An
interdisciplinary approach to long-term
modelling for power system expansion.
International Journal of Critical
Infrastructures, 2007. 3: p. 235-264.
Pruyt, E. System Dynamics Models of
Electrical Wind Power. in 22nd Conference of
the System Dynamics Society. 2004. Oxford:
International System Dynamics Society.
Chappin, E.J.L., G.P.J. Dijkema, and L.J. De
Vries, Carbon policies: do they deliver in the
Jong run?, in Caron Constrained: Future of
Electricity Generation, A. Press, Editor 2009,
Elsevier. p. 31-56.
European Climate Fund, Roadmap 2050, A
practical guide to a prosperous, low-carbon
Europe. Volume 1, Technical Analysis, 2010.
European Climate Fund, Roadmap 2050, A
practical guide to a prosperous, low-carbon
Europe. Volume 2, Policy Recommendations,
2010.
Del Rio Gonzalez, P., The interaction between
emissions trading and renewable electricity
support schemes. An overview of the
literature. Mitigation and Adaptation
Strategies for Global Change, 2007. 12(8): p.
1363-1390.
Held, A. et al., Re-Shaping: Shaping an
effective and efficient European renewable
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
15
energy market, 2010, Fraunhofer ISI &
Ecofys: Karlsruhe.
Verbruggen, A. and V. Lauber, Assessing the
performance of renewable electricity support
instruments. Energy Policy, 2012. 45: p. 635—
644.
Capgemini, Huropean
Observatory 2010, 2011.
Energy Markets
De Vries, L.J., The instability of competitive
energy-only electricity markets, in Research
Symposium Eurpoean Electricity
Markets2003, ECN: The Hague.
Agusdinata, B., Exploratory modeling and
analysis: A promising method to deal with
deep uncertainty, in Technology, Policy and
Management2008, Delft University of
Technology: Delft. p. 302.
Bankes, S., Exploratory modeling for policy
analysis. Operations Research, 1993. 41: p.
435-449.
Hamarat, C., J. Kwakkel, and E. Pruyt,
Adaptive Robust Design under Deep
Uncertainty. Technological Forecasting and
Social Change, 2012. Accepted for publication.
Pruyt, E. Using Small Models for Big Issues:
Exploratory System DYnamics Modelling and
Analysis for Insightful Crisis Management. in
18th International Conference of the System
Dynamics Society. 2010. Seoul, Korea.
Pruyt, E. and J.H. Kwakkel. A Bright Future
for System Dynamics: From Art to
Computational Science and Beyond. in The
30th International Conference of the System
Dynamics Society. 2012. St. Gallen,
Switzerland.
AEBIOM, E.B.A., 2011 Annual Statistical
Report on the contribution of Biomass to the
Energy System in the EU27, 2011: Brussels.
p. 1-108.
Boccard, N., Capacity factor of wind power
realized values vs. estimates. Energy Policy,
2009. 37: p. 2679-2688.
CIA. The World Factbook, Country
comparison: Electricity consumption. 2012
[cited 2012 30 May]; Available from:
https://www.cia.gov/library/publications/the-
world-factbook/rankorder/2042rank.html.
EEA. Total electricity consumption - Outlook
from IEA (Outlook 028) 2007 [cited 2012
August, 2); Available from:
http://www.eea.europa.eu/data-and-
maps/indicators/total-electricity-consumption-
16
42.
43.
44.
45.
46.
47.
48.
49.
outlook-from-iea/total-electricity-
consumption-outlook-from-1.
EEA. Passenger car ownership in the EEA.
2010 [cited 2012 9 August]; Available from:
http://www.eea.europa.eu/data-and-
maps/figures/passenger-car-ownership-in-the-
eea.
ENTSO-E, Indicative values for Net Transfer
Capacities (NTC) in Continental Europe,
2011: Brussels.
Eurostat. Share of renewables in gross inland
energy consumption, 2009. 2011 [cited 2011
18 December]; Available from:
http://epp.eurostat.ec.europa.eu/statistics_exp
lained/index.php?title=File:Share_of_renewab
les_in_gross_inland_energy_consumption,_20
09_(%25).png&filetimestamp=2011112410342
0.
Eurostat. Eurostat Energy Database. 2012
[cited 2012 6 March]; Available from:
http://epp.eurostat.ec.europa.eu/portal/page/po
rtal/energy/data/database.
He, Y.X., et al. Electricity demand price
elasticity in China based on computable
general equilibrium model analysis. Energy,
2011. 36: p. 1115-1123.
IEA, World Energy Outlook 2011 Factsheet,
2011: London.
IEA, NEA, and OECD, Projected Costs of
Generating Electricity, 2010: Paris, France.
Institute for
L
System (PVGIS)
Energy and_ Transport.
ical hi
formation
Geographical A of Solar Re
and Performance of Photovoltaic Technology.
2012 [cited 2012 30 May]; Available from:
http://re.jrc.ec.europa.eu/pygis/.
Junginger, M., W. Van Sark, and A. Faaij,
Technological Learning in the Energy Sector,
Lessons for Policy, Industry and Science2010,
Cheltenham: Edward Elgar Publishing Ltd.
Labandeira, X., J.M. Labeaga, and X. Lépez-
Otero, Estimation of elasticity price of
electricity with incomplete information.
Energy Economics, 2012. 34: p. 627-633.
Pezzulli, S., et al., The seasonal forecast of
electricity demand: A Hierarchical Bayesian
model with cli logical weather
Applied Stochastic Models in Business and
Industry, 2006. 22: p. 113-125.
Xiaoquing, H. and Y. Yong, Wind power
penetration limit calculation based on power
54.
56.
58.
59.
60.
61.
62.
63.
64.
system reliability, in Sustainable Power
Generation and Supply2009, IEEE: Nanjing.
Python. Python Programming Language
Official Website. 2012 25 February 2012 20
March 2012].
Parzen, E., On estimation of a probability
density function and mode The Annals of
Mathematical Statistics, 1962. 33(3): p. 1065-
1076.
Rosenblatt, M., Remarks on Some
Nonparametric Estimates of a Density
Function. The Annals of Mathematical
Statistics, 1956. 27(3): p. 832-837.
Kohavi, R. and G.H. John, Wrappers for
feature subset selection. Artificial
Intelligence, 1997. 97: p. 273-324.
Bryant, B.P. and R.J. Lempert, Thinking
inside the box: A participatory, computer-
assisted approach to scenario discovery.
Technological Forecasting & Social Change,
2010. 77: p. 34-49.
Forrester, J., System Dynamics, Systems
Thinking, and Soft OR. System Dynamics
Review, 1994. 10(2): p. 245-256.
Lehner, B., G. Czisch, and S. Vassolo. Model-
based assessment of European water
resources and hydrology in the face of global
change. 2001 [cited 2012 8 August]; Available
from: http://www. iset.uni-kassel.de/abt/w3-
w/projekte/europes_hydropower_bernhard.pdf.
World Energy Council, 2010 Survey of Energy
Resources, 2010: London, UK.
Bankes, S., R. Lempert, and S. Popper,
Making Computational Social Science
Effective. Social Science Computer Review,
2002. 20(4): p. 377-388.
Sterman, J.D., All Model are Wrong:
Reflections on Be ing a Systems Scientist.
System Dynamics Review, 2002. 18(4): p. 501-
531.
Meza, C.M.C. and G.P.J. Dijkema, Modelling
infrastructure systems: A hybrid approach for
system transition in First International
Conference on Infrastructure Systems and
Services: Building Networks for a Brighter
Future (INFRA)2008: Rotterdam.
